Dear colleagues:
For decades, certain qualitative techniques such as interpretive thematic analysis have worked in synergy with quantitative techniques such as bivariate analysis and multiple regression. In the last decades SEM has had a rapid growth among quantitative techniques. I would therefore like to know your opinion on these two questions:
1. How well does SEM fit into mixed-method studies?
2. In which mixed-methods designs does SEM work best?
Any kind of quantitative technique can fit into mixed methods if the design is well justified, and there is nothing special about SEM in this regard. For example, anytime you might use a regressions with survey data in mixed methods, SEM would just be a more advanced version of that.
Two mixed methods designs that might be a good fit with SEM are an exploratory sequential design (qual --> QUANT) and an explanatory sequential design (QUANT --> qual). In the first case, you could develop the content for a scale through a preliminary qualitative study and test that measurement structure through SEM. In the second case, if you had unexpected results from your SEM, you could use a follow-up qualitative study as a way to explain those results.
Any kind of quantitative technique can fit into mixed methods if the design is well justified, and there is nothing special about SEM in this regard. For example, anytime you might use a regressions with survey data in mixed methods, SEM would just be a more advanced version of that.
Two mixed methods designs that might be a good fit with SEM are an exploratory sequential design (qual --> QUANT) and an explanatory sequential design (QUANT --> qual). In the first case, you could develop the content for a scale through a preliminary qualitative study and test that measurement structure through SEM. In the second case, if you had unexpected results from your SEM, you could use a follow-up qualitative study as a way to explain those results.
Thank you very much David and Willem for your valuable suggestions.
Estimado Jorge,
Te cuento mi experiencia en el tema.
Recientemente realicé una investigaciones que inició con el método cualitativo, en donde, mediante la aplicación de entrevistas en profundidad y focus group se profundizó en la subjetividad de universitarios sobre la procrastinación y adicción al internet. Posteriormente, analicé la información lingüística mediante el proceso de codificación abierta, que permite identificar las categorías que emergen en las narrativas. Luego, con la codificación axial se identificó la categoría nuclear del fenómeno y con la codificación selectiva se propuso las relaciones existentes entre las categorías. Finalmente, mediante el procedimiento de elaboración del paradigma codificado, se realizó un esquema gráfico en donde se plasma la relación entre las categorías. Hasta este punto se cierra la investigación cualitativa y ya se obtiene un modelo hipotético que se puede testear en una siguiente fase cuantitativa.
En este punto, se construyó los instrumentos para valorar las variables observadas y se realizó su respectivo proceso de análisis de validez y confiabilidad.
Una vez que se contó con los instrumentos con propiedades psicométricas adecuadas, como lo es la consistencia interna, que permite eliminar los ítems que no cargan al coeficiente de Cronbach de la escala. Se aplicó a una muestra representativa que tuvo una relación de, al menos, 10 personas por cada ítem aplicado y se procedió a testear el modelo mediante el proceso de ecuaciones estructurales. En este paso se utilizó AMOS para SPSS en donde se puede dibujar el modelo obtenido en la fase cualitativa y se procedió a testearlo mediante SEM.
Espero haber aportado.
Saludos,
Carlos
Multi-trait, multi-method experiments,as described by Campbell & Fiske (1959) are almost exclusively limited to quantitative methods. In particular, the multi-trait, mult-imethod matrix that they describe relies on correlations, which do not apply to qualitative data.
Dear Carlos
Very interesting description you have made of your study. With all the details you have shared, it has been easy to understand how to use SEM in a mixed exploratory design.
Thank you very much.
For an elaborate discussion of the use of MTMM for evaluation of questions look at my publcations for example Saris and Gallhofer 2014.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling